chore: bit provider work

src/llmling_agent_providers/base.py

@@ -69,6 +69,7 @@ class AgentProvider[TDeps]:
     tool_used = Signal(ToolCallInfo)
     chunk_streamed = Signal(str, str)
     model_changed = Signal(object)  # Model | None
+    NAME: str
 
     def __init__(
         self,
@@ -198,3 +199,7 @@ async def get_token_limits(self) -> tokonomics.TokenLimits | None:
         except ValueError:
             logger.debug("Could not get token limits for model: %s", self.model_name)
             return None
+
+
+class AgentLLMProvider[TDeps](AgentProvider[TDeps]):
+    """Provider using LLM backend."""

src/llmling_agent_providers/callback.py

@@ -38,6 +38,8 @@ class CallbackProvider[TDeps](AgentProvider[TDeps]):
     - String or ChatMessage returns
     """
 
+    NAME = "callback"
+
     def __init__(
         self,
         callback: ProcessorCallback[Any],

src/llmling_agent_providers/human/provider.py

@@ -38,6 +38,7 @@ class HumanProvider(AgentProvider):
 
     model = None
     _conversation: ConversationManager
+    NAME = "human"
 
     def __init__(
         self,

src/llmling_agent_providers/litellm_provider.py

@@ -17,7 +17,7 @@
 from llmling_agent.models.content import BaseContent, Content
 from llmling_agent.models.messages import ChatMessage, TokenCost
 from llmling_agent_providers.base import (
-    AgentProvider,
+    AgentLLMProvider,
     ProviderResponse,
     StreamingResponseProtocol,
 )
@@ -108,9 +108,11 @@ def usage(self) -> Usage:
         return self._final_usage
 
 
-class LiteLLMProvider(AgentProvider[Any]):
+class LiteLLMProvider(AgentLLMProvider[Any]):
     """Provider using LiteLLM for model-agnostic completions."""
 
+    NAME = "litellm"
+
     def __init__(
         self,
         name: str,
@@ -260,7 +262,7 @@ async def generate_response(
                 tool_calls=calls,
                 model_name=model_name,
                 cost_and_usage=cost_and_usage,
-                provider_extra=response.choices[0].message.provider_specific_fields,
+                provider_extra=response.choices[0].message.provider_specific_fields,  # pyright: ignore
             )
 
         except Exception as e:

src/llmling_agent_providers/pydanticai/provider.py

@@ -26,7 +26,7 @@
     ToolSkippedError,
 )
 from llmling_agent.utils.inspection import has_argument_type
-from llmling_agent_providers.base import AgentProvider, ProviderResponse
+from llmling_agent_providers.base import AgentLLMProvider, ProviderResponse
 from llmling_agent_providers.pydanticai.utils import (
     convert_model_message,
     format_part,
@@ -49,10 +49,11 @@
 logger = get_logger(__name__)
 
 
-class PydanticAIProvider(AgentProvider):
+class PydanticAIProvider(AgentLLMProvider):
     """Provider using pydantic-ai as backend."""
 
     _conversation: ConversationManager
+    NAME = "pydantic_ai"
 
     def __init__(
         self,

src/llmling_agent_providers/remote_websocket/provider.py

@@ -65,6 +65,8 @@ class ToolContext(BaseModel):
 class WebSocketProvider(AgentProvider, TaskManagerMixin):
     """Provider that connects to remote agent via WebSocket."""
 
+    NAME = "websocket"
+
     def __init__(
         self,
         url: str,